You've just come back from a wedding. You took a hundred photos, and for each one you dictated an anecdote to KontexVoX: who was there, what was said, the exact moment. The descriptions are engraved in the EXIF. Now you want a physical album — not an online gallery nobody looks at, but a book you leaf through on the living-room coffee table, with well-written anecdotes next to each photo.

The challenge isn't a lack of material. It's sorting it. A hundred wedding photos means the ceremony, the speeches, the dancing, the stolen moments, the group shots, the details. Sorting all that by hand takes hours. But if every photo already carries a human description, an AI agent can do that sorting and storytelling work for you.

Prerequisites

Before you start, you need:

1. Photos enriched with KontexVoX. The descriptions live in each file's EXIF Description field. If you haven't enriched the photos yet, that's the step before this one — see What is an EXIF file?.
2. An AI agent with file-system access: Claude Code, Codex CLI, or Gemini. Not a web chatbot.
3. The photos exported to your computer via AirDrop or the Files app (the channels that preserve EXIF — see How to share without losing the EXIF).
EXIF doesn't survive an upload to a web chatbot. If you go through ChatGPT to generate your album, your descriptions are erased before the AI even reads them.

What you dictated: an example

To understand why the descriptions change everything, here's what the EXIF of four wedding photos contains:

IMG_0421.jpg → "This is the moment Marie sees her dress for the first time. Her mother is crying, her sister is telling her the lucky-doll tradition." IMG_0783.jpg → "Pierre's speech. He lost his notes, he's improvising, the room is laughing. That's when he said the line about the red sweater." IMG_1204.jpg → "Leaving the church, a rain of petals. Aunt Isabelle missed her rice throw — she always aims too far left." IMG_0956.jpg → "The first dance. Jean slipped, we thought he was going down, he caught himself laughing."

Those four lines, you said them out loud while looking at the photo. KontexVoX transcribed them and engraved them into the file. And it's precisely this human context that lets the AI do intelligent work: it reads "speech", "leaving the church", "first dance", and it understands the day's chronology without you having to explain it.

The prompt

Read all the JPEG images in this folder. For each image, extract the Description field from the EXIF metadata. From these descriptions, organize the photos into chronological chapters of the wedding day (getting ready, ceremony, speeches, dancing, stolen moments, details). For each chapter, write a short introduction that ties the anecdotes together in a warm, natural tone. Produce a layout document for a physical album: photo order, a rewritten and polished caption for each one, and accompanying text per chapter. Export everything as two files: a printable PDF and a summary CSV (page number, photo file, caption, chapter).

What the AI does with this prompt

The agent reads the EXIF descriptions, analyzes them, and deduces the structure of the day. It identifies the highlights, groups them into natural chapters, rewrites each anecdote as a polished caption (while keeping the original meaning and voice), and produces a layout document you can send to a photo-printing service or print yourself.

The result isn't a technical file nobody looks at. It's a physical album plan, ready to print, with the anecdotes of your day told as a story — not a flat caption.

The intelligent sorting is the heart of the prompt. Without the EXIF descriptions, the AI would have a hundred photos with no context. With them, it knows the first dance comes after the speeches, that getting ready comes before the ceremony, that "Aunt Isabelle" is a recurring character. It's the human context, engraved by voice, that makes this sorting possible.

Variation for another tool

With Codex CLI: same prompt, but also ask for a CSV export of the anecdotes so you can reuse them in a table. With Gemini: also ask for a continuous narrative that ties the anecdotes together like a literary story.

Frequently asked questions

Does it work if I have five hundred photos?

Yes, but the album gets long. Ask the agent to paginate (50 photos per page) or to generate multiple files per chapter of the day. The intelligent sorting works at any volume — it's the same classification job, just longer.

Why not an Apple Photos or Google Photos album?

Those albums exist and are excellent for sharing. But they don't showcase your anecdotes: they show photos sorted by date. The album generated by the AI agent puts your descriptions front and center, rewrites them as polished captions, and structures the day as a story. Above all, the result is printable: it's a physical object, not a screen.

Can I do the same thing without KontexVoX?

Technically yes — you can type the descriptions by hand into each photo, then run the same prompt. I did it, as a test: five photos is fine, twenty is tedious, fifty and you give up. KontexVoX replaces typing with your voice — that's the whole difference. The prompt itself works whatever the enrichment method.

Does the AI respect the meaning of my anecdotes?

Yes — the prompt asks it to rewrite and polish, not invent. Your descriptions are the source. The AI fixes the form (syntax, register) but keeps the meaning and the voice. If you want no rewriting at all, add "Keep the descriptions exact, do not rewrite them" to the prompt.

Going further

Building a slideshow for a best man's speech Generating a family story from contextualized photos Capture Mode or Classic Mode: which one for a wedding? Optimize an enriched-photo session Check that your photos are properly enriched Memories page — albums, weddings, travel